a16z shares World Labs' Atlas model for spatial intelligence from few photos
Posts highlight Atlas's new view prediction approach that unifies generation and reconstruction, cutting required photos for 3D capture from 100-300 down to three.
Posts highlight Atlas's new view prediction approach that unifies generation and reconstruction, cutting required photos for 3D capture from 100-300 down to three.
a16z posts quote World Labs co-founders on Atlas, built on new view prediction to unify pixel generation and reconstruction. The model reduces photos needed for 3D space capture by 50-100x, from 100-300 to three, and can recreate The Matrix bullet-time shot using three iPhones instead of hundreds of cameras.
96.6K
2 posts, first seen 4d ago
Posts highlight Atlas's new view prediction approach that unifies generation and reconstruction, cutting required photos for 3D capture from 100-300 down to three.
a16z posts quote World Labs co-founders on Atlas, built on new view prediction to unify pixel generation and reconstruction. The model reduces photos needed for 3D space capture by 50-100x, from 100-300 to three, and can recreate The Matrix bullet-time shot using three iPhones instead of hundreds of cameras.